CAPability: A Comprehensive Visual Caption Benchmark for Evaluating Both Correctness and Thoroughness
This work addresses the problem of evaluating detailed captions from multimodal large language models for researchers, though it is incremental as it builds on existing benchmarks by adding multi-view analysis.
The authors tackled the outdated evaluation of visual captioning by introducing CAPability, a comprehensive multi-view benchmark with 12 dimensions, which assesses both correctness and thoroughness using precision and hit metrics on nearly 11K annotated images and videos, revealing a performance gap between QA and caption capabilities.
Visual captioning benchmarks have become outdated with the emergence of modern multimodal large language models (MLLMs), as the brief ground-truth sentences and traditional metrics fail to assess detailed captions effectively. While recent benchmarks attempt to address this by focusing on keyword extraction or object-centric evaluation, they remain limited to vague-view or object-view analyses and incomplete visual element coverage. In this paper, we introduce CAPability, a comprehensive multi-view benchmark for evaluating visual captioning across 12 dimensions spanning six critical views. We curate nearly 11K human-annotated images and videos with visual element annotations to evaluate the generated captions. CAPability stably assesses both the correctness and thoroughness of captions with \textit{precision} and \textit{hit} metrics. By converting annotations to QA pairs, we further introduce a heuristic metric, \textit{know but cannot tell} ($K\bar{T}$), indicating a significant performance gap between QA and caption capabilities. Our work provides a holistic analysis of MLLMs' captioning abilities, as we identify their strengths and weaknesses across various dimensions, guiding future research to enhance specific aspects of their capabilities.